Open Bug 669684 Opened 13 years ago Updated 2 years ago

Avoid allocating a chunk for time-based GC trigger

Categories

(Core :: JavaScript Engine, defect)

x86
macOS
defect

Tracking

()

People

(Reporter: gwagner, Unassigned)

Details

(Keywords: memory-footprint, Whiteboard: [MemShrink:P3])

Right now we have to allocate a chunk in order to trigger the new time based GC. We could also trigger it when the last chunk with free arenas is picked and avoid a heap-growth. Should be another fragmentation improvement.
IIUC, the net effect is that we will avoid allocating a new chunk (and putting a single object into it just before doing GC) every time a time-based GC occurs.  Is that right?  Sounds like a small win, but a good one.
Whiteboard: [MemShrink]
(In reply to comment #1)
> IIUC, the net effect is that we will avoid allocating a new chunk (and
> putting a single object into it just before doing GC) every time a
> time-based GC occurs.  Is that right?  Sounds like a small win, but a good
> one.

We perform the GC when we return from executing a script so it's not like we allocate a chunk, put a single object in there and perform a GC right away.

But yes we can avoid allocating the new chunk in the low-allocation-rate situations.
Whiteboard: [MemShrink] → [MemShrink:P3]
Assignee: general → nobody
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.